Someone comments “GUIDE” under your Instagram post. They want the guide. Your first job is to get it to them without turning a small request into a sales interview.
That is a useful starting point for DM automation: fulfill an explicit request, make the next step easy, and give the person room to decide. A follow, a like, a conversation, and a sale mean different things. The flow should preserve those distinctions.
This guide focuses on a comment-to-DM campaign you can run, inspect, and improve. The goal is a useful conversation that can lead to business, with enough evidence to tell whether it did.
Choose an offer that makes the comment meaningful
“Comment YES for something exciting” tells you very little about what a person wants. “Comment FIT for the sizing guide for this jacket” tells you both the product and the problem. You can deliver an answer without guessing.
In Manychat’s interview with EZContacts’ head of growth, Rafael Sarim Oezdemir, he describes replacing any-comment triggers with specific keywords after the broader approach attracted low-intent contacts. That is a practitioner’s account published by a vendor, not proof that keywords will improve every campaign. It does suggest a good question: does your trigger reflect a request, or merely activity?
Start with one post and one offer. A buying guide suits someone comparing options. An availability link suits someone ready to book. A discount attracts interest in the discount, which may or may not become profitable demand. Choose the offer before writing the automation.
Leave ordinary conversation alone. Someone saying “great photo” has not asked to enter a qualification sequence.
Understand what the first DM permits
Meta’s private-reply documentation allows a private response to a qualifying comment. Follow-up messages require the recipient to respond and must fall within 24 hours of that response. A comment is therefore an entry point, not permission for an indefinite series of messages.
Manychat makes the distinction more specific: the opening private reply alone does not open the messaging window. A reply, regular button, or Quick Reply can establish the interaction needed to continue. An Open website button does not. See its comment-trigger documentation when you build the opening step.
You have two reasonable designs:
- Deliver the link immediately. Put the requested resource in the first private reply. This keeps delivery short. Accept that a website click alone does not let you continue the DM sequence.
- Offer an in-chat choice. Use a Quick Reply such as “Send the guide,” then deliver it after the tap. This creates an explicit conversational interaction. Make the button’s purpose obvious and avoid extra gates.
A like is not this comment trigger. Features for other entry points have their own eligibility and behavior; do not build a campaign around “message everyone who engaged.”
A small flow you can actually read
Here is an illustrative campaign for a shop selling a jacket in two fits. These messages are examples, not a report of client results.
- The post: Show how the regular and relaxed fits differ. End with “Comment FIT and we’ll DM you the size guide.”
- The opening DM: “Here for the jacket size guide? Tap below and I’ll send it.” Add a Quick Reply labeled “Send the guide.”
- The delivery: “Here’s the guide, including garment measurements and how to compare them with a jacket you own.” Link directly to it.
- The optional next step: “Want to see the regular fit, the relaxed fit, or ask us a question?” Route the choices to the relevant product page or a person.
The first message should identify itself as the shop’s automated helper if the presentation could otherwise imply a person is typing. Avoid simulated typing delays and claims such as “I’ve picked this especially for you” when the selection is just a fixed branch.
Notice what this flow leaves out: a required email address, an unrelated product pitch, and questions the product page already answers. Add a question only if its answer changes what you send next.
For a service business, replace the fit choice with the decision that matters for the service. Our real-estate workflow separates a request for property information from a request to speak with an agent.
Decide who handles the conversation after the buttons
People will type things you did not anticipate. “Will this arrive before Friday?” might require an inventory check and a shipping destination. A button labeled “Talk to us” is useful only when somebody owns that queue.
Assign a named team or person, set a response expectation you can meet, and pass along the campaign and choices already made. Pause the promotional sequence when a person takes over. The customer should not receive a discount nudge while an employee is explaining why the order cannot ship.
Keep answers about price, stock, delivery, or appointments tied to a maintained source. If the automation cannot check a changing fact reliably, send the relevant page or route the question. Confidently repeating yesterday’s stock count is worse than saying an employee will check.
Choose tools around this operating model. Manychat documents the comment-to-DM flow in detail. If you evaluate a team messaging platform such as respond.io, test the handoff and channel setup as carefully as the first response. Its Instagram documentation also makes clear that the human-agent extension is not permission for automated follow-ups.
Follow up while it is useful and allowed
A delayed message needs both a reason and an eligible messaging window. “Did that guide answer your sizing question?” might help someone who asked for advice. Sending it to everyone who opened a link creates noise.
Manychat says it prevents scheduled messages from sending outside the standard 24-hour window. Build with that behavior in mind. A three-day nurture sequence in your flow editor does not establish permission for three days of Instagram messages.
For longer-term communication, offer an email subscription separately and explain what it contains. Let people receive the promised resource without silently adding them to unrelated marketing. Respect a refusal or stop request across the campaign.
Test the awkward paths before judging the copy
Test from an account that has never messaged the business, including a nonfollower. Meta says private replies to nonfollowers arrive in Requests. A message that was sent may not have been noticed.
Also test the opening Quick Reply, the direct website link, a free-text question, and an unavailable product. Manychat’s comment trigger only activates for the first comment a person leaves on a post or reel; repeatedly commenting from your usual test account can give you a misleading failure.
A Manychat community discussion describes an operator whose test accounts worked while other accounts did not. It is anecdotal troubleshooting, not a failure-rate study. Its practical lesson is useful: a successful preview is insufficient evidence that the published route works for a new visitor.
Measure the handoffs, then change one thing
Record how many distinct people requested the resource, interacted with the opening DM, reached the destination, and completed the business action. Keep sent messages and website clicks separate from purchases or confirmed appointments. Count the latter in the system that actually records them.
Use a campaign-specific link so downstream activity can be connected to the flow where measurement permits. Report missing attribution honestly; a cross-device purchase or a shared link can break a neat person-by-person funnel. A campaign code can add evidence, but it does not prove the DM caused the sale.
If requests are healthy and few people continue, inspect the opening message and Requests behavior. If people reach the product page but do not buy, examine the offer, stock, price, and checkout. More DM copy may be the wrong repair.
Review real conversations alongside the counts. Start with one recurring problem, fix it, and compare the next set of conversations. The DM Campaign Field Guide can help you write down the offer, owner, and completion event before building a larger flow.
First published . Updated with current research and practical examples.
